[Dev Catch Up #42] - Llama 3.3-70B, Chain-of-Continuous-Thought Paradigm, Valkey-based Caching, GenAI for Parsing Web Pages, React Data Fetching Patterns, and much more.

Bringing devs up to speed on the latest dev news from the trends including, a bunch of exciting developments and articles.

Welcome to the 42nd edition of DevShorts, Dev Catch Up!

I write about developer stories and open source, partly from my work and experience interacting with people all over the globe.

Thanks for reading Dev Shorts! Subscribe for free to receive new posts and support my work.

Some recent issues from Dev Catch up:

• Llama 3.2, Molmo Multimodal Models, Infinite Queries

• Microsoft's Drasi and bitnet.cpp, next-safe-action library

• Qwen2.5, xLAM-1b, TanStack Router, Nvidia's Minitron SLMs

• OpenAI's Realtime API, Deep-learning for ETA

Join 2000+ developers to hear stories from Open source and technology

Must Read

• Within a year of Llama’s launch, Meta announced its new and most advanced language model Llama 3.3. It is an auto-regressive language model that uses an optimized transformer architecture. This pre-trained instruction tuned generative model trained on 70B parameters is optimized for multilingual dialogue use cases. Know more about this model from its huggingface page.

• LLMs are restricted to reason in the language space where they process the reasoning process with a Chain-of-Thought to solve a complex problem. But this is not optimal for reasoning. Hence, this new paradigm Chain-of-Continuous-Thought is introduced which enables LLMs to reason in the latent space. Learn more about this from here.

• Interaction to Next Paint or INP is a performance metric that measures the time from a user interaction to the next visual change rendered on the screen. This article guides you in depth about the patterns that a developer can use to improve INP when using React.

• Code-renderers determine how UI elements are transformed into something with which an user can see or interact with. Now, if these renderers are not able to handle files with large amount of code, the application will ultimately crash. Here is an article that guides you to build a code renderer with the help of virtualisation that can handle large amount of code and coverage data.

After this, shouting out a top open-source project is a delight and here it is:

OSS Highlight of the Week

• This week the open-source project that really caught the attention of the community is VectorChord. It is a PostgreSQL extension and designed for scalable and high-performance vector similarity search. Check it out from its GitHub page and leave a star to support it.

  

Now, we will head over to some of the news and articles that will be at a place of interest for developers and the tech community out there.

Hope you are enjoying this edition of our newsletter so far! Support us by giving a free follow to our LinkedIn and X pages.

Your support is highly appreciated!

Good to know

• Next.js helps you to build Server-Side Rendered applications but faces significant challenges when dealing with them. The best way to mitigate the blocking nature of this applications is to cache the maximum amount of possible data. This article guides you on how you can do with the highly effective Valkey caching solution.

• With GenAI continuously evolving, it is now possible to use the technology to convert a web page into structured data. This article tells you exactly how you can do it.

• Language agents can be supplied with different mechanisms in enormous quantity to accomplish autonomous tasks. But current mechanisms follow a fixed order that limits the ability of the agents. However, this paper proposes Activation Learning with Self-Exploration, which focuses on optimizing mechanism activation adaptability without reliance on expert models.

• Normalizing Flows or NFs are likelihood-based model for continuous inputs but got little attention with all the developments happening around AI. With the aim of showing how powerful NFs are, this paper presents an architecture named Tarflow that enables highly performant NF models.

Lastly, we will take a look at some of the trending scoops that hold a special mention for the community.

Notable FYIs

• Code agents are the new hot things that are taking developer productivity through the roof. This podcast gives you a complete picture on building production coding agents with the CEOs of Stacblitz and Qodo along with insights on flow engineering.

• Infusing AI capabilities into Java Applications is possible with LangChain4j and its Quarkus integration. Learn more about it from this short article with slides that guides you how to leverage LLMs into one.

• While SQL can perform CRUD operations and enforces data integrity, it faces significant challenge to integrate with languages like JavaScript and Python because of its composable programming techniques. This article guides you how to write composable SQL using Knex and pipelines.

• While building react applications, data fetching is a common task. Here is an article that discusses the different patterns for fetching data in react components.

• Measuring a product’s availability and latency helps in maintaining and enhancing the reliability of that product. SLOs are used for these metrics. Know how these SLOs are kept up-to-date with E2E testing in a microservices architecture from this detailed article.

That’s it from us with this edition. We hope you are going away with a ton of new information. Lastly, share this newsletter with your colleagues and pals if you find it valuable and a subscription to the newsletter will be awesome if you are reading for the first time.